The Worldwide Journal of Robotics Analysis, Forward of Print.
The exploration of large-scale unknown environments can profit from the deployment of a number of robots for collaborative mapping. Every robotic explores a piece of the atmosphere and communicates onboard pose estimates and maps to a central server to construct an optimized world multi-robot map. Naturally, inconsistencies can come up between onboard and server estimates because of onboard odometry drift, failures, or degeneracies. The mapping server can right and overcome such failure instances utilizing computationally costly operations equivalent to inter-robot loop closure detection and multi-modal mapping. Nonetheless, the person robots don’t profit from the collaborative map if the mapping server supplies no suggestions. Though server updates from the multi-robot map can vastly alleviate the robotic mission strategically, most present work lacks them, because of their related computational and bandwidth-related prices. Motivated by this problem, this paper proposes a novel collaborative mapping framework that allows world mapping consistency amongst robots and the mapping server. Specifically, we suggest graph spectral evaluation, at totally different spatial scales, to detect structural variations between robotic and server graphs, and to generate essential constraints for the person robotic pose graphs. Our strategy particularly finds the nodes that correspond to the drift’s origin slightly than the nodes the place the error turns into too giant. We completely analyze and validate our proposed framework utilizing a number of real-world multi-robot discipline deployments the place we present enhancements of the onboard system as much as 90% and may get better the onboard estimation from localization failures and even from the degeneracies inside its estimation.
发布者：Lukas Bernreiter，转转请注明出处：https://robotalks.cn/a-framework-for-collaborative-multi-robot-mapping-using-spectral-graph-wavelets/
